Physics-informed Neural Networks to Model and Control Robots: a Theoretical and Experimental Investigation

TL;DR

This paper explores the use of physics-informed neural networks for modeling and controlling complex robots, extending the approach to handle non-conservative effects, and validating the methods through real-world experiments.

Contribution

It introduces extensions to physics-informed neural networks for non-conservative effects and combines them with traditional controllers, providing theoretical stability and experimental validation.

Findings

Achieved precise control with stability guarantees.

Validated models on soft robot motion prediction.

Demonstrated trajectory tracking on a robotic manipulator.

Abstract

This work concerns the application of physics-informed neural networks to the modeling and control of complex robotic systems. Achieving this goal required extending Physics Informed Neural Networks to handle non-conservative effects. We propose to combine these learned models with model-based controllers originally developed with first-principle models in mind. By combining standard and new techniques, we can achieve precise control performance while proving theoretical stability bounds. These validations include real-world experiments of motion prediction with a soft robot and of trajectory tracking with a Franka Emika manipulator.